We parametrize the multiplicative analysis of the total James map given by Caruso and ourselves. We prove that if the given coefficient system e has a sum, the product of the targets is a ring space. This is only a fragment of the full structure present in the most interesting examples.
We parametrize the multiplicative analysis of the total James map given by Caruso and ourselves. We prove that if the given coefficient system e has a sum, the product of the targets is a ring space. This is only a fragment of the full structure present in the most interesting examples.
We parametrize the multiplicative analysis of the total James map given by Caruso and ourselves. We prove that if the given coefficient system e has a sum, the product of the targets is a ring space. This is only a fragment of the full structure present in the most interesting examples.
Volume 2RI, Number I, January 19R4 JAMES MAPS AND En RING SPACES BY F. R. COHEN, J. P. MAY AND L. R. TAYLOR ABSTRACT. We parametrize by operad actions the multiplicative analysis of the total James map given by Caruso and ourselves. The target of the total James map ) == C(R n , X) IT QDq(R n , X)
is an En ring space and) is a en-map, where en is the little n-cubes operad. This implies that) has an n-fold delooping with domain X. It also implies an algorithm for the calculation of)* and thus of each ()q)* on mod p homology. When n == 00 and p == 2, this algorithm is the essential starting point for Kuhn's proof of the Whitehead conjecture. In [2, 4], we analyzed the behavior with respect to pairings of the lames maps i q : ex QDq(e, X) used in [4] to obtain the stable splittings of spaces ex. In particular, we proved that if the given coefficient system e has a sum, then the product of the targets is a ring space and the product of the iq's is an exponential H-map. In fact, this is only a fragment of the full structure present in the most interesting examples. In the first two sections, we give a parametrized elaboration of this analysis. We assume the given coefficient system e is a "module" over an operad and we deduce, essentially, that (jq) : ex IT QD q {e, x)
is a -map. Here denotes the weak infinite product (all but finitely many coordinates of each point are the basepoint). Technically, we shall replace the target by its equivalent Q( V D q ( e, X)) and by an equivalent, but larger, operad. This analysis requires the introduction of general notions of additive and multi- plicative actions of operads on coefficient systems, and these generalizations of the definitions of an operad and of an action of one operad on another give considerable added flexibility to the theory of iterated loop spaces developed in [9-13]. When eis e(R n ), our results imply that (jq) has an n-fold delooping IT QDq(W,X)).
Received by the editors January 20, 1983. 1980 Mathematics Subject Classification. Primary 55P35, 55P47, 55S12; Secondary 55Q25. Key words and phrases. James maps, En ring space, operad, coefficient system, homology operations. 1984 American Mathematical Society 0002-9947/84 $1.00 + $.25 per page 285 286 F. R. COHEN AND J. P. MAY AND L. R. TAYLOR In the case n == 2, this can be improved to a 2-fold delooping IT C(R 2 q,D q (R 2 ,X))),
and C(R 2 q, D q (R 2 , X)) may be replaced by X) if X is path con- nected. Moreover, the sequence of maps i q : C(R n , X) QDq(R n , X) is entirely determined by the multiplicative structure of the "En ring space" QDq(R n , X). This fact implies an algorithm for the calculation of the maps i q on homology. We discuss these implications in 3. Kuhn [6] has made extensive mod 2 calculations on the basis of the cited algorithm, and he has recently used these calculations to construct a homotopical resolution of the sphere spectrum [7], thus generalizing the Kahn-Priddy Theorem and proving a long-standing conjecture of G. W. Whitehead. Ralph, Cohen has pointed out that the spectrum level analogs of the basic constructions ex and Dq(e, X) can be used to obtain a totally noncombinatorial conceptual construction of the space level James maps and to prove their multiplica- tive properties. The main ideas are explained in his paper [5]. We originally planned to say more about this here but have decided to defer the discussion to [8], where the foundational work necessary to make it rigorous is carried out. This is a sequel to [2], but only 2-4 of that paper are essential preliminaries to this one. 1. Additive actions of operads on coefficient systems. We here relate the canonical James maps of separated coefficient systems to operad actions. As we shall see in a moment, the following is a natural generalization of the definition of an operad [9, p. 1]. DEFINITION 1.1. Let be an operad and ea coeffcient system. Then eis said to be a -module if there are maps y: k X e. X ... Xe. e., il ik i which satisfy the following properties. (a) If g E k' h r E jr for 1 r k, and Cs E e is for 1 s then y( y(g; hI' ... ,h k ); Cl' ,Cj) == y(g; d l , ,d k ), where d r == y(h r ; c jl + ... +jr-I+ I ' ,c jl + .. '+jr) with dr == * ifir == o. (b) y(l; c) == c, where 1 E I is the unit of . (c) If g E k' c r E e jr , and 0 E then y ( go; Cl'. . . ,Ck) == y ( g; Ca-I (l) , . . . 'Ca-I ( k) ) 0 ( i I , . . . ,ik ) , where o(il' ... ,ik) E is the permutation (J j V ... Vjk V ... Vja-I(k) the isomorphisms being block sum identifications. (d) For each fixed g E k' the maps y(g) specify a k-fold sum on e, that is, a natural transformation e k e0 V, where V : A k A is the k-fold wedge. JAMES MAPS AND En RING SPACES 287 DEFINITION 1.2. A -module eis said to be separated if eis separated and if each sum y( g) is separated in the sense that the maps y(g) : IT 8. X ... Xe 8. II lk 1 )1+ ... +jk=j induced by passage to orbit spaces are all injections; eis said to be weakly separated if each y(g) is weakly separated in the sense that these orbit maps induce a k-fold pairing on {e(8 j )}; compare [2,4.1]. EXAMPLE 1.3. When e== , only the right unit formula y(g; I k ) == g needs to be added to make Definition 1.1 coincide with the definition of an operad. In particular, an operad is a -module. The little cubes operad en is a separated en-module. EXAMPLE 1.4. Let == en and let e== e(Jn X Y) for any space Y, where J is the open unit interval. Then e is a separated -module with respect to the maps y specified as follows. For a little cube I: In In and an element b == <z1' .. , Zj) of F(J n X Y, ))withz s == (X s ' Ys)' X s E In andys E Y, define fb == (lxI' YI)'' (Ixj , Then define y(g; CI''C k ) == (gICI,,gkCk)E F(J n X Y,)t + ... +)k) for g == <gI' ... ,gk) E en k and c r E F(J n X Y, )r)' 1 r k. These maps y are precisely analogous to the structural maps y of en' and centerpoint projection specifies morphisms' gn : en e( In) of en-modules. Henceforth we write R instead of J, by abuse. EXAMPLE 1.5. If e is a -module and e' is a '-module, then e X e' is a X '-module. If e or e' is weakly separated, then so is e X e'. (The analogous assertion with "weakly separated" replaced by "separated" is false.) Recall from [9, p. 5] that an action of an operad on a space X with basepoint 1 is a sequence of maps k X X k X such that *) == 1, x) == x, and y(g; hI' ,hk); Xl' 'Xj) == YI)' Yk)) for g E k' h r E jr and X s E X, where) ==)1 + ... +)k andYr is the rth block of)r variables x s A -space with zero, or o-space for short, is a -space with a second basepoint 0 such that Xl' ... ,X k ) == 0 if any X r == O. Thus induces a based map D r (, X) X, where the smash powers of X are taken with respect to the basepoint o. We need a number of structures implied by Definitions 1.1 and 1.2. First, we have the following generalization of the fact that ex is a e.space if eis an operad [9, 2]. LEMMA 1.6. Let ebe a -module. For based spaces X, ex is lunctorially a -space. PROOF. For g E k' c r E ejr' andYr E Xjr,.define k( g; [c1; Y1] , , [ck; Yk]) == [ y ( g; Cl' .. , Ck); Y1, ... ,Yk] . The requisite verifications are trivial. 288 F. R. COHEN AND J. P. MAY AND L. R. TAYLOR LEMMA 1.7. Let ebe a 9-module. For a based space X, define D( e, X) == V D q ( e, X). q"?O Then D(t', X) is functorially a 9 o -space. PROOF. Do(e, X) == SO, the wedge uses the basepoint 0, and the point 1 is to be the unit for the -action. The maps in the previous proof pass to quotients to give Lk -equivariant maps 9: 1\ D j1 ( e, X) 1\ ... 1\ D jk ( e, X) D j ( e, X), i == it + ... +ik' and these provide the required action. LEMMA 1.8. Let ebe a weakly separated g-module and define 8 == Il 8 q , where 8 q == eq/L q with 80 == {*}. q"?O For 9 0 -spaces X, C( 8, X) is functorially a go-space. PROOF. C( 8, X) is defined using the basepoint 0 E X, and the unit of C( e, X) is the element [<*); 1]. For g E gk' the action of g is the "composite" ( f3 )[k] ((jk [k]) C(y(g), (Cl ) Cc,X Cc,X Cc,X, where Q9 is given by [2,1.9]. The weak separation property is just enough to ensure that the second map is defined on the image of the first; it is only defined on the entire middle space when eis a separated g-module. Continuity in g is clear, and the requisite associativity, unity, and equivariance formulas are inherited from those of y and the When X == V X q , the external sum EB : V C(8 q , X q ) C(8, X) of [2, 2.5] is an inclusion. Its image consists of those points [< Cl' ... ,ck ); Xl' ... ,Xk] which are diagonally paired in the sense that c r E 8 q if and only if x r E X q , for the same q. Provided that g) carries JSk 1\ ... 1\ JSk to JS, i == il + ... +ik' this subspace is closed under the 9-action. On the domain of EB, 9k may then be viewed as a parameter space indexing k-fold multiplications D of the sort introduced in [2, 2]. We have chosen to work in the larger space for conceptual clarity and for technical convenience in the next section. When e is a separated coefficient system, we agree to regard the following composite as the canonical total lames mapi: IT c(e q , D q (2, X)) D(e, X)); q"?O compare [2, 4]. When e == e(Jn),i is induced by sending a finite subset of In to its power set (see [2, 6]) and is therefore exponential. The following basic observation is then just a parametrization of this fact by little cubes. PROPOSITION 1.9. Let e be a weak(y separated g-module. Then i: CX C(8, D(t', X)) is a map ofg-spaces. lAMES MAPS AND En RING SPACES PROOF. It suffices to verify that the following diagram commutes for g E k: 289 j
t c(e k , D(e, xi k ]), tc( y(g), c(e, D(e, X)) The results of [2, 2-4] admit evident generalizations to k-fold pairings, and our diagram is just that of [2, 3.4] as obtained from the k-fold weakly separated sum y(g): e k e. Thus it commutes by [2, 4.2]. While we shall generally use this result in the form given, we can of course restrict the target to the domain of EB. This is advantageous in the following example. EXAMPLE 1.10. Consider e(R n ) c C3'(R n ), where C3'(R n )q == Rnq. While C3'(R n ) is of course not separated, it does satisfy the injectivity condition required of separated en-modules in Definition 1.2. Therefore Lemma 1.8 applies to it and we have maps of en-spaces: (Jq) C(R n , X) IT C(B(R n , q), Dq(R n , X)) IT C(R n q/"2: q , Dq(R n , X)).
Take n == 2. Here R 2 q/"2: q C q /"2: q cq R 2 q by the fundamental theorem of algebra. This implies a "delooping" X IT C(R 2 q,D q (R 2 ,X))).
(See the discussion in 3.) 2. Multiplicative actions of operads on coefficient systems. We want to transport Proposition 1.9 along the chain of maps connecting its target to QD( e, X). For this purpose, we reinterpret the action of on the target in terms of the following definition. DEFINITION 2.1. Let be an operad and ea coefficient system. Then is said to act multiplicatively on e if -there is a unit element 1 Eel and there are maps A: k X eJI X ... xe Jk e J ,} ==}l .. }k' which satisfy the following properties. (a) If g E k' h r E Jr for 1 r k, and Cs E e zs for 1 S + ... +}k' then A(y(g; hl, ... ,h k ), Cl, ... ,CJI+"'+Jk) == A(g; dl, ... ,d k ), where d r == A(h r ; c JI + ... +Jr-I+l' ... ,C JI +. +Jr) with d r == 1 if}r == o. (b) A(I, c) == C for C E e J and A(g; l k ) == 1 for g E k. (c) If g E k' c r E Jr and (J E "2: k , then A( g(J; Cl' . , Ck) == A( g; C(J -I (1) , , C(J -I (k) ) (J <)1, ,}k) , 290 F. R. COHEN AND J. P. MAY AND L. R. TAYLOR where (J <it, ... ,ik> E is the permutation (J j jt 1\ ... I\jk ja-I(l) 1\ ... I\ja- I (k) j, the isomorphisms being lexicographic identifications. (d) For each fixed g E k' the maps A(g) specify a k-fold product on 8, that is, a natural transformation 8 k 8 0 1\, where 1\ : A k A is the k-fold smash product. EXAMPLE 2.2. When 8 is also an operad, only the analog of (a) involving the structural maps y of 8 needs to be added to make this definition coincide with the definition of an action of an operad on an operad 8 ([11, VI, 1.6], with misprints corrected in [13, 1.8]). Thus the operad pairs (8, ) used in [11] and [13] provide examples. The following example is implicit in Lemma 1.8. LEMMA 2.3. Let 8 be a weakly separated -module. Then acts multiplicatively on the configuration space coefficient system 8( 8). PROOF. The point * E 8 0 c 8 == F(8, 1) is taken to be the unit of 8(8). For g E k' the maps A(g) are the "composites" ( -) (-) 0 (-k ) F(y(g),j) (- ) F 8, it x XF 8, ik 8 ,i )F 8, i , i == it ...ik The requisite verifications are straightforward. Now the action of on e(8, X) of Lemma 1.8 is seen to be a special case of the following generic kind of action (which is a direct generalization of [11, VI.l.8]). LEMMA 2.4. Let act multiplicatively on 8 and let X be a o-space. Then ex is functorially a o-space and the natural map TJ : X ex is a o-map. PROOF. ex is defined using the basepoint 0 E X, TJ: X ex is specified by TJ( x) == [1; x] and the zero and unit for ex are the elements TJ(O) == [*; *] and TJ( 1). The action on exis specified in terms of Aand the action on X by the formula for g E k' c r E 8 J r' and Yr == (xr,t, ... ,xr,Jr) E XJr, where I runs over the lexico- graphically ordered set of sequences (i t, ... , ik) with 1 ir and where y/ == (Xt,i l ' ,Xk,i k ) E X k The requisite verifications are elementary exercises. The following omnibus result, whose unexplained notations will be clarified in the course of the ensuing discussion, will be used in conjunction with Proposition 1.9. PROPOSITION 2.5. Let 8 be a weakly separated -module and X a o-space. Then all of the following are X e-maps: - 171 _ - 172 lX oo e(8, X) e(8, X) QX. Here eis the linear isometries operad of [11, p. 10], so that e J is the space of linear isometries (Roo)J R oo . It is an E oo operad; that is, e J is and contractible. In multiplicative infinite loop space theory, the little cubes operad 8 00 used in [2] is an JAMES MAPS AND En RING SPACES 291 inappropriate choice for the canonical additive E oo operad. The little convex bodies operad of [11] was a clumsy first approximation to the right choice. The definitive choice was introduced by Steiner [14]. It is denoted %00' and K oo X denotes the associated construction on spaces. As in the proof of [2, 4.6], C( 8, X) is constructed from the product coefficient system 8(8) == 8(8) X %00' ?T 1 and ?T 2 are induced from the projections, and ?T 1 is a homotopy equivalence. Of course, X e acts multiplicatively on 8( 8) by pullback along the projection X e . Similarly, X e acts on the space X. It is a basic property of %00 that e, and therefore also X e, acts on it. We conclude that X e acts on 8(8) in such a way that the projections to 8( 8) and to %00 commute with the actions. Therefore ?T1 and ?T 2 are maps of X e-spaces. It is another. basic property of %00 that there is a natural map X oo : K oo X QX which is a homotopy equivalence if X is connected and a group in general. Moreover, QX is a X e-space and X oo is a X e-map when X is a o-space. These facts imply the proposition. Actually, much more is true; (%00' X e) is an operad pair and K oo X is the free (%00' X e)-space generated by the o-space X; QX is also a (%00' X e)-space and X oo : KooX QX is a (%00' X e)-map. Intuitively, these statements mean that K oo X and QX are H-semiring spaces with additions and multiplications parame- trized by %00 and X e. The parameter spaces encode higher homotopies, and X oo preserves all of this information. When == 8 n , a (%00' X e)-space is said to be an En ring space. When is an E oo operad, a (%00' X e)-space is said to be an E oo ring space. The between the additive and multiplicative operad actions provides a very powerful tool for the analysis of the latter. We refer to [11-14] for details on all of this and turn to a discussion of its calculational significance in our context of lames maps. 3. Implications, interpretations, and delooping theorems. Let be an operad, 8 a weakly separated -module, and X a space with, basepoint *. Consider the following commutative diagram:
ex ---------X QD(8, X) (A) _ j 1 __ 1 2 W2 iU oo e(8,D(8, X)) e(8,b(8, X)) ) K oo D(8, x) Here 11 is the based inclusion, 11( x) == [1; x], determined by a unit 1 E 8 1 Clearly 11 induces a based inclusion, still denoted 11, of X in D 1 ( 8, X). If 8 1 is contractible, as always holds in practice, the latter inclusion is a homotopy equivalence. Consider the composites D q (8, X)). For q == 0, 8 0 == {*} and lo collapses all of ex to the point [<*), 1], where 1 E SO == D o ( 8, X); compare [2, 3]. For q == 1, direct calculation from the defini- tions gives 292 F. R. COHEN AND J. P. MAY AND L. R. TAYLOR Similarly, for q > 1,}q1J is constant at the basepoint. Thus we must define == (}1J)(x) ==[ Ef) 0 ==[(*,1); 1, 1J(x)]. If x == *, this is the same point as the unit [<*); 1] of the -space C( 8, D( e, X)). To define and simply replace <*, I)E F(8,2) on the right by *, 1), k 2 ) and k 2 , where k 2 is any chosen point of %00,2' Of course, == Note that is connected to the unit by a path, but is not equal to the unit, if i > 1. Indeed, as discussed in [9, p. 4], a path in %00,1 connecting y(k 2 ; *, 1) to the unit 1 determines all of the required paths. If, as we assume, * is a nondegenerate basepoint of X, we can replace 1J and the by homotopic maps such that the diagram is still strictly commutative but the new maps all carry * to the unit of their target X e-spaces. To see this, let X' == X V I with new basepoint at the end of the whisker I. Extend 1J and over the whisker by use of the constant path and extend i > 1, by use of the paths just indicated. There results a commutative based diagram with domain X'. The projection X' X is a based homotopy equivalence, and precomposition with a chosen homotopy inverse gives the required new maps 1J and Let (G X L)( X) denote the free X e-space generated by X. Since, by Proposi- tions 1.9 and 2.5,}, 17'1' 17'2' and a oo are all X e-maps, there results a commutative diagram of X e-spaces and X e-maps: Choosing inverses to equivalences, let} also denote the composite total lames map When e== is an operad, or when e== e(R n ) and == en' ij is an equivalence [4, 1.7]. In this case, j is homotopic to f 4 1J-1. Here f 4 is entirely determined by its restriction to X, which is completely understood, and by the X e-structure on the (%00' X e)-space QD(e, X). To see the calculational significance of this assertion, consider mod p homology for a prime p. If eis an E oo operad, then H*CX is a functor of H*X. More precisely, H*CX is functorially generated by H*X via the Pontryagin product and homology operations determined by the action of eon CX. See [3, 1.4.1]. We think of this structure as additive. With e== , we think of the resulting -action on D( e, X) of Lemma 1.7 as multiplicative. The induced multiplicative Pontryagin product and homology operations on H*D( e, X) can be read off by passage to filtration quotients from the additive structure on H*CX. Precisely analogous assertions, but with a more complicated set of homology operations involved, are valid when e== en == . See [3, 111.3.1]. Now consider QX. Again H*QX is functorially generated by H*X via the (additive) Pontryagin product and homology operations determined by the action of %00' See [3, 1.4.2]. For any E oo ring space, the general formulas of [3, 11, 1-3] determine the multiplicative Pontryagin product and homology operations on ele- ments decomposable under the additive Pontryagin product and homology opera- tions. Although the details have yet to be worked out, it is clear from the geometric JAMES MAPS AND En RING SPACES 293 diagrams on hand that such formulas must also exist for En ring spaces. When X is a o-space, QX is a (%00' X e)-space; when is 8 n or an E oo operad, we conclude that the multiplicative Pontryagin product and homology operations on H*QX are algorithmically determined from this structure on H*X. See [3, 11, 4] for a more detailed discussion. Returning to QD( 8, X), we conclude that if 8 == is 8 n or an E oo operad, then the multiplicative Pontryagin product and homology operations on H*QD( 8, X) are algorithmically determined by H*X. The total James map J: CX QD(8, X) converts the additive structure on H*CX to the multiplicative structure on H*QD(8, X), and we know the restriction of J* to H*X. We therefore have an algorithm for the global computation of J*, To compute the qth James map on homology, we first compute the total map and then project to the qth coordinate via D(8, X) X). To see what is going on more clearly, it is useful to redescribe H*QD( 8, X) in terms of the composite equivalence IT q Qi q * IT QD q ( 8, X) IT QD( 8, X) QD( 8, X),
where i q : D q (8, X) D(8, X) is the inclusion and * is the loop sum. Up to homotopy, the map of diagram (A) is the composite of this equivalence and the map X X) with coordinates (1,11,0, ... ,0, ... ), where 1 denotes the . 11 constant map at i E SO C QSo, 11 denotes the composIte X X) C QD 1 (8, X), and the remaining coordinates are constant at 0. Since we are using weak infinite products, H*( IT QD q (8, X)) = Q9H*QD q (8, X).
In H*QD( 8, X), the tensor products here become loop sums, in line with the fact that H*QD(8, X) is generated under the additive Pontryagin product and homology operations by H*D( 8, X) == H*D q ( 8, X). ' We refer to Kuhn [6, 7] for concrete calculations and geometric applications based on these ideas and turn to homotopical, rather than homological, implications of our study of operad actions. If 8 n X 8 acts on a space X, then there is an n-fold classifying space BnX == C n X L, X) and a natural map t: X which is a weak equivalence if X is connected and a group completion in general (if n 2). See [9, 13] and, for the nonconnected case, [10, 2] and [3, p. 487, # 11]. In our theory we often begin with multiplicative actions by 8 n X e on ring spaces X. Unless we shrink to appropriate unit components, X will be trivial (due to the presence of zero). Identify QD q (8, X) with the unit subspace of QD q (8, X), namely the subspace consisting of those points whose Oth coordinate is 1. A compatible embedding of Q(V D q (8, X)) in QD(8, X) is obtained by mapping a point/to 294 F. R. COHEN ANDJ. P. MAYANDL. R. TAYLOR the loop sum 1 * (Qi)(f), where i: V q ~ l D q (8, X) ~ D(8, X) is the natural based inclusion and 1 is the unit element. We write UQD(e, X) ~ Q( V Dq(e, X)) q ~ l for this unit space. Upon restriction to unit subspaces of the targets of the maps f i , we can apply the functor En to the entire diagram (B) when == en. In particular, inverting equiva- lences, we obtain a map Bl: B J1 CX ~ B n UQD(8, X) and a homotopy commutative diagram Thus} becomes an n-fold loop map upon group completion. If X is connected, the maps L are equivalences and we conclude that} itself is an n-fold loop map. When == 8 00 , or any other 00 operad, we can carry out precisely the same discussion with n== '00. Here Boo is to be interpreted as the spectrum-valued functor of [9, 14], and goo is to be interpreted as the Oth space functor from spectra to spaces. In this case,} is an infinite loop map if X is connected. At this writing, these facts are more curious than useful, and further investigation is certainly called for. One would like to identify the spaces involved, in particular BnCX and gnBnUQD(e, X). For the latter, we have the following weakly homotopy commutative diagram (at least if ?TaX is countable): X QD q (8, X) ..--------- gnB n UQD(8, X) q ~ l Here the left arrow is the inclusion of the weak infinite product in the full infinite product. We proved in [2, 1.7] that the latter is a grouplike multiplicative H-space, and [2, 1.2] gives the dotted arrow. It seems reasonable to conjecture that the dotted arrow is a weak equivalence. If X is connected, then D q ( e, X) is (q - 1)-connected and all maps in the diagram are weak equivalences. In general, verification of the conjecture would require better understanding of the homology of infinite products than we possess. We can identify BnCX in many interesting cases. Ife == en' then BflCX is equivalent to ~ n x by [9, 13.1(vii)]. Of course, this conclusion carries over to the case 8 ==8(R"). Much more generally, Caruso [1] has proven that BnC(R n X Y, X) is equivalent to C( Y, ~ n X) for any weak metric space Y. lAMES MAPS AND En RING SPACES BIBLIOGRAPHY 295 1. J. Caruso, Configuration spaces and mapping spaces, Thesis, Univ. of Chicago, 1979. 2. J. Caruso, F. R. Cohen, J. P. May and L. R. Taylor, James maps, Segal maps, and the Kahn-Priddy theorem,Trans. Amer. Math. Soc. 281(1984), 243-283. 3. F. R. Cohen, T. Lada and J. P. May, The homology of iterated loop spaces, Lectures Notes in Math., vol. 533, Springer-Verlag, Berlin and New York, 1976. 4. F. R. Cohen, J. P. May and L. R. Taylor, Splitting of certain spaces CX, Math. Proc. Cambridge P h i l o s ~ Soc. 84 (1978), 465-496. 5. R. L. Cohen, Stable proofs of stable splittings, Math. Proc. Cambridge Philos. Soc. 88 (1980), 149-152. 6. N. J. Kuhn, The homology of the James-Hopf maps, Illinois 1. Math. 27 (1983), 315-333. 7. , A Kahn-Priddy sequence and a conjecture of G. W. Whitehead, Math. Proc. Cambridge Philos. Soc. 92 (1982), 467-483. 8. L. G. Lewis, Jr., J. P. May, 1. McClure and M. Steinberger, Equivariant stable homotopy theory, Lecture Notes in Math. (to appear). 9. J. P. May, The geometry of iterated loop spaces, Lecture Notes in Math., vol. 271, Springer-Verlag, Berlin and New York, 1972. 10. , E oo spaces, group completions and permutative categories, London Math. Soc. Lecture Notes Ser., no. 11, Cambridge Univ. Press, Cambridge, 1974, pp. 61-93. 11. , (with contributions by F. Quinn, N. Ray and J. Tomehave), E oo ring spaces and E oo ring spectra, Lecture Notes in Math., vol. 577, Springer-Verlag, Berlin and New York, 1977. 12. __, Pairings of categories and spectra, J. Pure Appl. Algebra 19 (1980),199-346. 13. __, Multiplicative infinite loop space theory, J. Pure Appl. Algebra 26 (1982), 1-69. 14. R. Steiner, A canonical operad pair, Math. Proc. Cambridge Philos. Soc. 86 (1979), 443-449. DEPARTMENT OF MATHEMATICS, UNIVERSITY OF KENTUCKY, LEXINGTON, KENTUCKY 40506 DEPARTMENT OF MATHEMATICS, UNIVERSITY OF CHICAGO, CHICAGO, ILLINOIS 60637 DEPARTMENT OF MATHEMATICS, UNIVERSITY OF NOTRE DAME, NOTRE DAME, INDIANA 46556